High Performance Cooling of Traction Brushless Machines

(1)

THESIS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

High Performance Cooling of Traction Brushless

Machines

ALESSANDRO ACQUAVIVA

Department of Electrical Engineering Division of Electric Power Engineering CHALMERS UNIVERSITY OF TECHNOLOGY

(2)

High Performance Cooling of Traction Brushless Machines ALESSANDRO ACQUAVIVA

ISBN 978-91-7905-433-5

c

ALESSANDRO ACQUAVIVA, 2021.

Doktorsavhandlingar vid Chalmers Tekniska Högskola Ny serie nr. 4900

ISSN 0346-718X

Department of Electrical Engineering Division of Electric Power Engineering Chalmers University of Technology SE–412 96 Göteborg

Sweden

Telephone +46 (0)735704363

Chalmers Bibliotek, Reproservice Gothenburg, Sweden 2021

(3)

(4)

(5)

High Performance Cooling of Traction Brushless Machines ALESSANDRO ACQUAVIVA

Department of Electrical Engineering Chalmers University of Technology

Abstract

The work presented in this thesis covers several aspects of traction electric drive system design. Particular attention is given to the traction electrical machine with focus on the cooling solution, thermal modelling and testing.

A 60 kW peak power traction machine is designed to achieve high power density and high efficiency thanks to direct oil cooling. The machine selected has a tooth coil winding, also defined as non-overlapping fractional slot concentrated winding. This winding concept is state of the art for many applications with high volumes and powers below 10 kW. Also, these have been proven successful in high power applications such as wind power generators. In this thesis, it is shown that this technology is promising also for traction machines and, with some suggested design solutions, can present certain unique advantages when it comes to manufacturing and cooling.

The traction machine in this work is designed for a small two-seater electric vehicle but could as well be used in a parallel hybrid. The proposed solution has the advantage of having a simple winding design and of integrating the cooling within the stator slot and core. A prototype of the machine has been built and tested, showing that the machine can operate with current densities of up to 35 A/mm2_{for 30 seconds and 25 A/mm}2_{continuously. This results in a net power}

density of the built prototype of 24 kW/l and a gross power density of 8 kW/l with a peak efficiency above 94%. It is shown that a version of the same design optimized for mass manufacturing has the potential of having a gross power density of 15.5 kW/l which would be comparable with the best in class traction machines found on the automotive industry.

The cooling solution proposed is resulting in significantly lower winding temperature and an efficiency gain between 1.5% and 3.5% points, depending on the drivecycle, compared to an external jacket cooling, which is a common solution for traction motors.

(6)

(7)

Acknowledgements

First and foremost, I would like to express my gratitude to my main supervisor Prof. Sonja Lundmark and my examiner Prof. Torbjörn Thiringer for their great support, guidance and mentoring. I would also like to thank my co-supervisor Dr. Emma Grunditz for her valuable time, kindness and contribution. I really enjoyed working with all of you.

A special thanks to Dr. Stefan Skoog for sharing the daily challenges and for the very fruitful collaboration. I am grateful also to all the other colleagues and friends at Electric Power Engineering for creating such a great work environment. I had the chance and pleasure to spend some time at KTH with Prof. Wallmark and at Politecnico di Torino with Prof. Guglielmi, these experiences enriched my knowledge and my experience as a researcher.

Finally, I would like to thank my family for unconditional love and always supporting my choices no matter how far from home I get. I am infinitely grateful to my dad for inspiring and challenging me as a kid and over the years.

I gratefully acknowledge The Swedish Energy Agency for the financial support.

Alessandro Acquaviva Gothenburg, Sweden February, 2021

(8)

(9)

4.1 Set-up 1 - Main investigated machine with closed loop oil-to-water circuit . . . 45 4.2 Set-up 2 - Electrical machine open loop water circuit . . . 50 4.3 Set-up 3 - Power converter open loop water circuit . . . 53 5 Electric drivetrain thermal modelling and analysis 55 5.1 System modelling . . . 55 5.2 System analysis . . . 58 5.3 Comparison - water/glycol cooling jacket vs direct oil cooling . . . 62

6 Conclusions and future work 67

Bibliography 71

References . . . 71

Appendices 76

A Assumptions and derivations for Table 1.1 77

B LP network derivations 81

B.1 Direct cooling machine LP . . . 81 B.2 Cooling jacket machine LP . . . 83

(11)

List of Symbols

j imaginary unit -d differential operator -v voltage [V] i current [A] R resistance [Ω] L inductance [H] l incremental inductance [H] Ψ flux linkage [Wb] p number of poles

-ω electrical angular frequency [rad/s] n mechanical angular speed [rpm]

ξ saliency -Cth thermal capacitance [J/K] Rth thermal resistance [K/W] T torque [Nm] Θ temperature [K] V volume [m3_] ˙

V volumetric flow rate [m3_/s]

A area [m2_]

r radius [m]

θ mechanical angle [rad]

ρm specific mass [kg/m3]

Cp specific heat [J/(kg K)]

λ thermal conductivity [W/(m K)]

(12)

Chapter 0 Contents

µr relative permeability

-µ0 permeability of vacuum [H/m]

(13)

Introduction

1.1 Problem background

One of the challenges of the twenty-first century is fighting climate change. En-vironmental research highlights how there is an increase in the average Earth temperature, which is significantly faster than temperature swings registered in the past, and the trend is not promising. According to scientists in the field, a raise of 2◦C would cause a significant and irreversible climate impact. Glacier melting and an raising sea level are just part of the possible consequences. Research has correlated the temperature increase with the greenhouse gas emission increase. Another aspect is that city pollution is a concern for health and quality of life.

Passenger vehicles are responsible for around 12% (data from 2014) of total EU emissions of carbon dioxide (CO2), the main greenhouse gas. If vans and heavy

duty vehicles are included the number rises above 20% [1]. The transportation impact on the greenhouse gas emission is significant, as can be seen in Fig. 1.1.

With these premises, and assuming that at the source there is a shift to renewable energy, a move towards a sustainable transportation is a clear need to avoid climate change. The trend is positive, in fact, the market for electric vehicles (EV) is strong, with an annual market growth rate above 40% year-on-year from 2010 [3].

As of today, the biggest technological challenges in the EV industry, are related to the energy storage and charging infrastructure. However, the electric drivetrain efficiency, compactness and cost play a very important role in the development of the future generation of EVs. In a few years from now, traction electric drivetrains will be produced with the rate of millions per year [3]. Ensuring

(14)

Chapter 1 Introduction 1990 1995 2000 2005 2010 2015 2020 year 0 20 40 60 billion tons

Greenhouse gas emissions World

Transportation Electric energy and heat Total 1990 1995 2000 2005 2010 2015 2020 year 0 1 2 3 4 billion tons

Greenhouse gas emissions EU

1990 1995 2000 2005 2010 2015 2020 year 15 20 25 [%] Transportation % share EU World

Figure 1.1: Greenhouse gas emissions in the World and in EU. Greenhouse gas emissions are measured in tonnes of carbon dioxide-equivalents (CO2e) [2].

a simple, effective and reliable manufacturing process of both the power converters and traction electrical machines is the key to succeed in the electric vehicle industry. It took many years to reach the current quality for the manufacturing process of the internal combustion engine. However, a much faster development is needed for the electric drive-train if the sustainability goals, set for 2030, about CO2

emissions are to be met. EU aims at reducing the net human-caused emissions of carbon dioxide (CO2) by at least 55 percent from the 2010 levels by 2030, reaching

(15)

1.2. TECHNOLOGICAL BACKGROUND

‘net zero’ around 2050 [1].

1.2 Technological background

The industry trend is to move towards highly integrated drivetrains, where often the electric machine and the power electronics share the same cooling loop, aiming at high power density, high efficiency and low cost.

Semiconductor technology development is one of the factors that can enable high power density converters. The introduction of wide band-gap devices opens new possibilities to improve the performance of the traction inverter. The thermal capability and the low switching losses of silicon carbide (SiC) MOSFETs can be beneficial in comparison to classical silicon (Si) IGBTs when used in a three-phase converter [4–7] allowing a more compact design.

High-power density electric machines can be achieved by minimizing the non-active parts, such as the end-windings, and by design parameters, in particular • high airgap flux density (i.e. magnetic loading), eg. by using high energy magnets in combination with core material with saturation at high flux density [8]

• high mechanical speed [9]

• high current loading, or high current density, while simultaneously assuring a low thermal resistance between the winding and coolant

For the sake of high magnetic loading, the development of electric drive-trains is dominated by the permanent magnet synchronous machine (PMSM) [10], characterized by its high power density as well as high efficiency [11–13]. The use of rare earth magnets, such as NdFeB, in combination with flux concentration rotor structures, such as the V-shaped interior permanent magnet (IPM) rotor, allows to reach high peak airgap flux densities (up to 1 T) [11, 12]. Industrial core materials can operate typically up to 1.6-1.8 T, limited by high saturation and losses. There are materials, such as cobalt-iron, that enable higher flux densities, up to 2-2.2 T. However, higher cost and iron losses limits the use of these materials to aerospace and niche applications such as motor-sports [14]. As of today, with the current available materials, there is an upper limit when it comes to magnetic loading.

Rotating electrical machines generated output torque is proportional to the size, assuming the same current and magnetic loading condition [11, 12]. This

(16)

Chapter 1 Introduction

means that the output power and the mechanical rotating speed, for a fixed size machine, are linearly related. So it is convenient, from a specific power perspective, defined as power per weight ratio, to aim for high mechanical rotational speed. The upper limit is usually given by the yield strength of the rotor materials, the bearings capability, the core and copper AC losses which increase with the electrical frequency as well as the ability of the converter to operate at very high frequencies [15, 16]. Furthermore, a gearbox is required to match the required wheel rotational speed with the traction machine rotor speed. The higher the machine speed, the higher the needed gear ratio (number of gear stages) which can add complexity and cost to the drivetrain. Typically, maximum speeds for traction machines are in the range of 12000-18000 rpm [17].

To reach even higher power density, as identified in [18], the machine needs to be able to withstand high current density. A lot of research and engineering efforts are dedicated to “improved thermal materials”, as well as “advanced cooling/thermal management techniques to reduce size, cost and improve reliability” as summarized well in [19–21]. An apparent aim is to try to bring the coolant medium closer to the main sources of heat, i.e. the stator core and winding, as opposed to so called cooling jackets. One interesting development comes from high thermally conductive epoxy potting materials. These are used for example in the rail industry to improve heat conduction and electrical isolation as well as to increase the mechanical stiffness (and damp vibrations) [22].

Enabling high current density, which in turn means high joule losses in the winding, at peak operation, does not prevent the traction machine from achieving high energy efficiency. The reason is that during driving, the machine operates most of the time in part-load, i.e. the low torque region, as shown for several drive cycles in [23] and in Chapter 5. Achieving high energy efficiency can significantly extend the driving range of the vehicle for a given battery pack.

Electrical machine windings can be divided in two main categories, dis-tributed windings (DW) and non-overlapping fractional slot concentrated windings (FSCW) also referred as tooth-coil winding machines (TCWMs). With TCWMs, it may be possible to devote some of the space in the slot that is normally used for active material (conductors), for cooling channels instead, without sacrificing per-formance. Consequently, it may offer high torque density and high efficiency [24–26] when combined with a permanent magnet (PM) rotor, as well as low manufacturing cost. In [27], a 12-slot 8-pole interior-magnet TCWM for traction applications is compared with a distributed winding, a switched reluctance, and an induction

(17)

1.2. TECHNOLOGICAL BACKGROUND

machine. The TCWM is shown to perform best in terms of torque density, even without considering the shorter winding overhang. Efficiency-wise, the two PM machines are comparable, TCWM being slightly better in the low speed region and less efficient at high speeds compared to the DW machine.

Permanent magnet synchronous machines are found in most electric vehicles available on the market today. An overview of the main specifications of some of the traction machines found in the automotive industry, which can be considered state-of-the-art, are presented in Table 1.1.

Table 1.1: Overview and performance for traction machines in automotive industry

Tesla BMW Toyota Chevy Audi

Model 3 i3 Prius IV Bolt A3 e-tron

Year 2017 2016 2017 2016 2014 Peak torque [Nm] 348 250 163 360 330 Peak power [kW] 202 127 53 150 75 Base speed [rpm] 4800 4500 3400 3500 -Max speed [rpm] 18100 11400 17000 8810 6000 Active volume [l] 5.32 6.35 2.21 4.11 5.56 Gross volume [l] 12.7* 14 4.7* 8.7* 10.3*

Type of cooling Jacket, shaft Jacket Jacket Jacket Jacket

Type of winding Distributed Distributed Hairpin Hairpin Concentrated

Net power density [kW/l] 38 20 24 36.5 13.5

Gross power density [kW/l] 16* 9.1 11.3* 17.2* 7.3*

Net torque density [Nm/l] 65.5 39.4 73.8 87.6 59.4

Gross torque density [Nm/l] 27.5* 17.9 34.9* 41.4* 32.1*

Reference [28] [28, 29] [28] [28, 30] [28]

*These values are derived based on the assumptions presented in Appendix A

The machines presented in Table 1.1 are all radial flux machines and present different types of winding, including the hairpin winding, a type of distributed winding. The power densities, both net (calculated considering active length and the outer stator diameter) and gross (volume of complete electric machine including the non active parts and the housing) are shown in Table 1.1. These have been specified using the volume and not the weight (the power to weight ratio is defined as specific power) due to the uncertainty in the weight data found in the literature. Furthermore, typically, the weight to volume ratio does not vary much for radial flux machines. The Tesla Model 3 and Chevy Bolt machines are the ones presenting the highest power densities. It is important to highlight that there is a degree of uncertainty in the comparison due to the fact that it is not stated under which conditions the peak torque and peak power values are valid. The Chevy Bolt machine peak torque can be held for about 25 s for a 70◦C temperature rise [30], the coolant flow rate is unknown. No such information could

(18)

be found for the other machines in Table 1.1. By looking at the net torque density, and assuming that the magnetic loading is not so different among the machines analyzed, it can be noticed that the hairpin machines achieve the highest values thanks to the high copper space factors achieved. These machines, however, are typically limited in mechanical speed due to AC losses in conductors. The Tesla machine combines good values of torque density with a high base and maximum mechanical speed.

The aim of the machine design presented in this work, when engineered for mass production, is to reach values of gross power densities comparable to the ones shown in Table 1.1, thus comparable to the best in class found in the EV industry.

1.3 Review of previous work

Some interesting traction motor designs using TCWMs are presented in [31–35], however, without the integration of direct cooling in the stator, continuous current densities above 20 A/mm2 _{are hardly reached, which limits the torque density,}

and in turn the power density.

Several proposals of high performing direct cooling techniques for TCWMs can be found in the literature.

• A double layer tooth coil winding machine concept with in-slot cooling between the coils is presented and partly evaluated in [36]. This solution uses the space in the slot not filled with copper to create cooling channels by using water-soluble mould cores, a concept that is hard to adopt for mass production.

• Using conductive pipes in the slots, with the drawback of generating large eddy current losses [37]

• Theoretical evaluation of the concept of flushing the entire stator and rotor with oil coolant [38]

• Direct-water cooled coils by winding a coolant carrying steel pipe with Litz wire, validated in a 205 kW machine for a bus application [39]. The prototype presented uses Litz wire with a tube for liquid inside each turn, which is complicated to manufacture and yet the maximum feasible current density 14 A/mm2 _{at 2 l/min is reported.}

(19)

1.4. PURPOSE OF THE THESIS AND CONTRIBUTIONS

layer tooth coil would machine in [40]. Current densities of 25 A/mm2

continuous and 40 A/mm2 peak operation are reported with this solution, with coolant flow rates up to 5.3 l/min and 5.1 kPa pressure drop. Copper heat exchangers in the slot, however, can be challenging when it comes to slot insulation and manufacturing, and the authors have not presented a complete rotating machine in hardware with their cooling concept.

• The authors of [41, 42] present an in-slot cooling for a switched reluctance machine using a fluid guiding structure and airgap sealing to allow for oil cooling within the slots. The concept is tested with DC current up to 22 A/mm2and a flow rate of 6 l/min, however this concept comes with some challenges regarding coolant leakage to the rotor.

An emerging electrical machine technology, which has proven very effective in terms of power density, is the yokeless axial flux machines combined with liquid cooling. In particular YASA and Magnax have been showing solutions with gross power densities of over 25 kW/l (10 kW/kg) [17]. These numbers are impressive and are as of now the highest values found for commercial products within the traction power industry. However, an issue of the yokeless axial flux machine with surface mounted PMs is the low field weakening capability which typically forces an oversizing of the electric machine.

1.4 Purpose of the thesis and contributions

A cooling solution combining

• in-slot cooling, using the new available high thermally conductive potting materials to create the cooling channels

• direct iron cooling, by having stator yoke cooling channels

• design of the end section to properly distribute the oil flow to form a high turbulence region which leads to very high cooling capability of the end windings

is missing in the literature. Furthermore, the solutions found in the literature are often interesting concepts at a prototype level but hardly implementable with a low cost manufacturing process.

The main purpose of the work presented in this thesis is to investigate the different aspects of the multiphysics modelling of brushless PM machines and present a specific design of a traction machine with high cooling capabilities meant

(20)

for mass production. The targets of the design are to achieve high efficiency and power density compared to the state of the art. The focus is on radial flux tooth coil winding machines with integrated direct cooling. A novel solution is proposed, built and tested showing promising results in terms of torque and power density and a good agreement with the performance predicted both by analytical equations and numerical methods. The solution is presented in Paper VII, Paper IX, Paper X and in Chapter 3 of this work.

Other related activities are conducted and presented. A simplified numerical approach to modelling of cooling jackets for electrical machines is validated through thermal measurements and presented in Paper V. Also, a MOSFET inverter conduction loss prediction model is presented and validated experimentally with a SiC MOSFET inverter in Paper VIII. Finally, a system model of the converter, machine and the cooling circuit is built to analyze the performance of the system at different coolant flow rates and driving conditions. The model is also used to compare the cooling solution proposed in this work with an external cooling jacket in Chapter 5.

Below follows a more detailed description of the main contributions: • A strategy to select pole and slot combinations from both a bottom-up (from

performance evaluation of pole slot combinations) and top-down (from the specifications) design criteria for double layer tooth coil winding machines is proposed. (Paper II )

• A procedure based on analytical expressions to size brushless AC PM ma-chines based on split ratio is proposed, quantifying how with constant current density there is a clear trade-off between torque density and efficiency, with the cost optimum usually laying in-between. (Paper III )

• A novel cooling design for tooth coil winding machines is presented, analyzed and validated experimentally. The novelty consists in the integration of the cooling within the stator, using a thermally conductive epoxy resin to create the channels within the slot, the positioning of the stator yoke cooling channels as well as the design of the end section to cool the end windings. (Paper VII, Paper IX )

• It is shown how the manufacturability of a machine such as the one mentioned in the previous point can be improved with the use of a linear winding machine that pre-winds the coils on a bobbin. This could potentially lead to a reduced manufacturing cost for high volume production, depending on

(21)

1.5. THESIS OUTLINE

the automation level of the process. The various steps of the manufacturing process are analyzed and improvements to make the machine mass producible are proposed. (Paper X )

• The implementation of three different custom made calorimetric set-ups to validate the efficiency and losses is presented. The measurement system solution and components adopted can be valuable for other researchers in the field. (Paper V, Paper VII, Paper VIII, Paper IX )

• A simplified numerical approach to the analysis of electrical machines with cooling jackets is presented and validated experimentally. (Paper V ) • A quantification and derivation of the energy efficiency, with focus on

con-duction losses and the consequence of reverse concon-duction for a MOSFET inverter is presented and verified using a three-phase SiC MOSFET inverter suitable for traction applications. (Paper I, Paper VIII )

• A system level comparison of the cooling solution proposed and an exter-nal cooling jacket, using experimentally validated convection heat transfer coefficients, is performed in terms of efficiency and internal machine temper-atures varying the flow rate of the coolant and with different drive-cycles. (Chapter 5)

1.5 Thesis outline

The thesis is structured as a collection of articles, which means that the main part of the work is found in the papers attached at the end of the thesis. The thesis includes a clarification concerning how they are interrelated, some additional design aspects which are not found in the papers, a drivetrain cooling system study and a summary of the results.

In Chapter 2, the scientific and industrial context of the traction electrical machine drive technology is discussed with a particular focus on cooling and tooth coil winding machines. Furthermore some relevant theory about PMSM and thermal modelling is outlined.

Chapter 3 outlines the design and sizing of a tooth coil winding PM machine with a novel cooling concept which has been built and tested. Starting from the vehicle performance requirements all the way to the manufacturing details.

Chapter 4 covers the three calorimetric set-ups which have been used for experimental validation of the analytical and numerical models for both machines

(22)

and MOSFET inverter, summarizing the main results achieved.

Chapter 5, presents a system model and analysis of the electric drivetrain with a particular focus on thermal and cooling aspects. The system model is built with lumped parameter thermal models and loss maps and is used to evaluate performance of different cooling strategies depending on the driving conditions.

Chapter 6 presents the main conclusions regarding the work presented and suggested future work.

1.6 List of publications

This thesis is based on the work contained in the following papers. A contribution statement is added when the author of this work is not the first author and main contributor of the papers listed.

Paper I - A. Acquaviva and T. Thiringer - Energy efficiency of a SiC MOSFET propulsion inverter accounting for the MOSFET’s reverse conduction and the blanking time. Published in European Conference on Power Electronics and Applications (EPE’17 ECCE Europe), Sept. 2017.

Paper II - S. Skoog and A. Acquaviva - Pole-Slot Selection Considerations for Double Layer Three-phase Tooth-Coil Wound Electrical Machines. Published in International Conference on Electrical Machines (ICEM), Sept. 2018. Contribution to the idea and structure of the paper as well as the manuscript writing.

Paper III - A. Acquaviva - Analytical Electromagnetic Sizing of Inner Rotor Brushless PM Machines Based on Split Ratio Optimization. Published in International Conference on Electrical Machines (ICEM), Sept. 2018.

Paper IV - S. Lundmark, A. Acquaviva and A. Berqvist - Coupled 3-D Thermal and Electromagnetic Modelling of a Liquid-cooled Transverse Flux Traction Motor. Published in International Conference on Electrical Machines (ICEM), Sept. 2018.

Minor contribution being involved in the review and part of the writing.

Paper V - A. Acquaviva, O. Wallmark, S. Lundmark, E. Grunditz and T. Thiringer - Computationally Efficient Modeling of Electrical Machines with Cooling

(23)

1.6. LIST OF PUBLICATIONS

Jacket. Published in IEEE Transactions on Transportation Electrification, Vol. 5, Issue 3, p. 618-629, Sept. 2019.

Paper VI - A. Acquaviva, E. Grunditz, S. Lundmark and T. Thiringer - Comparison of MTPA and Minimum Loss Control for Tooth Coil Winding PMSM Consid-ering PM and Inverter Losses. Published in European Conference on Power Electronics and Applications (EPE’19 ECCE Europe), Sept. 2019.

Paper VII - A. Acquaviva, S. Skoog and T. Thiringer - Design and Verification of In-slot Oil-Cooled Tooth Coil Winding PM Machine for Traction Application. Published in IEEE Transactions on Industrial Electronics, Early Access, 2020.

Paper VIII - A. Acquaviva, A. Rodionov, A. Kersten, T. Thiringer and Y. Liu - Analytical Conduction Loss Calculation of a MOSFET Three-Phase Inverter Accounting for the Reverse Conduction and the Blanking Time. Published in IEEE Transactions on Industrial Electronics, Early Access, 2020.

Paper IX - A. Acquaviva, S. Skoog, E. Grunditz and T. Thiringer - Electromagnetic and Calorimetric Validation of Direct Oil Cooled Tooth Coil Winding PM Machine for Traction Application. Published in Energies, Vol. 13, Issue 13, 3339, 2020.

Paper X - A. Acquaviva, S. Skoog and T. Thiringer - Manufacturing of tooth coil winding PM machines with in-slot oil cooling. Published in International Conference on Electrical Machines (ICEM), Aug. 2020.

Paper XI - A. Rodionov, A. Acquaviva and Y. Liu - Sizing and energy efficiency analysis of a multi-phase FSCW PMSM drive for traction application. Published in IECON20, Oct. 2020.

Contribution with the machine model, analysis of results and part of manuscript writing.

(24)

(25)

Electrical machines state of the

art and modelling

The traction electrical machine is where the electromechanical energy conversion takes place and ideally this must happen in the most efficient way, occupying the least space (and consequently being as light as possible), it needs to last for the whole life of the vehicle and the overall cost and environmental impact should be as low as possible. Additionally, other features are required, such as low noise and vibrations, easiness in the control and in the recycling of materials. The design process aims at finding the best trade-off among several of these requirements for the specific application.

2.1 Traction electrical machines state of the art

Traction electric drives can be categorized in numerous ways, for instance by the electrical machines in use, the number of machines per vehicle and the placement of the traction machines within the vehicle. Regarding the traction machine, the ones mainly found are induction machines and PM (or brushless) synchronous machines. Currently there is no technology that is proven best. The brushless synchronous machine typically presents higher efficiency, power factor and power density compared to the induction machine. In the latter, losses generated in the rotor by the induced currents are a substantial part of the total losses, affecting directly the efficiency and often requires special rotor cooling. However, the efficiency of the induction machine can improve significantly with copper rotor bars, shown in Fig. 2.1. The main advantage of the induction machines comes from not having the permanent magnets, which otherwise when used in other

(26)

Chapter 2 Electrical machines state of the art and modelling

machine types introduce some challenges such as:

• Risk of demagnetization at high rotor temperatures

• Electromotive force (EMF) unregulated over-voltage in case of a fault • Substantial iron losses (power loss in the steel laminations) during coasting

and in the low torque region

• High cost of the rare earth materials

• Environmental impact of rare earths mining and processing • Significant cost in the mounting process

• Depending on the speed special magnet retention techniques might be re-quired

Figure 2.1: A cut open induction machine with copper rotor bars.

The most common machine type in the traction automotive industry is the brushless ac machine [10], which can be further divided by the rotor type and by the winding type. The rotor type presents many alternatives but mainly interior PM and PM assisted synchronous reluctance rotors are used. As mentioned in the introduction chapter, for the winding types there are two main alternatives, distributed winding and fractional slot concentrated windings, also defined as a tooth coil winding. Machines with these two types of windings are comparable in performance.

(27)

2.1. TRACTION ELECTRICAL MACHINES STATE OF THE ART

One type of distributed winding which is currently state of the art in many traction applications is the hairpin winding, shown in Fig. 2.2. This presents a simplified manufacturing process by insertion and good copper space factors, also known as fill factor. However, this technology is limited only to a low number of turns (pins) per stator slot and it requires welding at the end connections of the pins which complicates the manufacturing. Furthermore, there are problems of ac losses in the conductors at high rotational speeds, i.e. at high frequencies.

Figure 2.2: Hairpin winding with welded and potted end connections on one side.

2.1.1 Tooth coil winding machine

Tooth-coil wound machines, also known as non-overlapped fractional slot concen-trated winding (FSCW) synchronous machines, offer several benefits compared to machines with a distributed winding, but they also present some special charac-teristics resulting in design concerns not typical for classical distributed winding machines. The tooth coil winding presents the following advantages:

• Short end windings

• Low noise and torque ripple [24, 25] (this depends very much on pole slot combination)

(28)

• High slot copper space factor achievable, especially if combined with a segmented stator (such as plug-in teeth or joint-lapped core) [24, 25] • High values of stator leakage inductance [43, 44]. A high value of d-axis

inductance enables to de-flux the machine with less current

• Reduced fault vulnerability due to lower mutual inductance between phases [43] While the main drawbacks are:

• The winding factor which is proportional to the Torque/Ampere ratio is lower compared to the DW machine [45]

• Space harmonics generate additional losses in the rotor core and PMs, these can be significant, especially at high speeds [24, 25]

• Some pole/slot combinations can produce unbalanced magnetic forces [46] • Sub-harmonics may cause low dominant vibration modes which may result

in high acoustic noise and vibration

• It is typically hard to reach high saliency values due to high values of leakage and harmonic inductance. This usually limits the ability to have a significant amount of reluctance torque, which becomes very important in the field weakening region

When it comes to the production process, tooth coil winding machines present several options. A pole chain stator is shown in Fig. 2.3 and a segmented teeth stator is shown in Fig. 2.4. Some of these options, such as the pole star with yoke ring, single poles with yoke ring and the stator without pole shoes, present the opportunity of pre-winding the coils on bobbins and inserting them. This can reduce greatly the production time and cost by using a spindle winding machine. There are mainly three types of winding machines:

• Spindle winding or linear winding machine • Flyer winding machine

• Needle winding machine

A cost analysis of the winding manufacturing process and assembly is very impor-tant during the design of the machine and is dependant on the number of units per year to be produced. The higher the number of units, the more it is worth investing in a fully automated process and designing a machine that can fit the process. The variables are many to be considered, to mention some: number of

(29)

2.1. TRACTION ELECTRICAL MACHINES STATE OF THE ART

Figure 2.3: Pole chain stator.

Figure 2.4: Segmented stator, the stator teeth and yoke are put together after inserting the preformed coil.

coils per motor, achievable copper space factor, number of internal coil connections, wire diameter and number of parallel strands. The needle winding machine, for example, allows to wind the stator coils of the same phase without disconnecting the wire and can easily manage to wind machines that have coils connected in parallel. With the spindle winding machine, a coil is produced by winding the wire on a rotating body at very high speeds and many coils can be produced in parallel with a single machine. An overview of motor winding technology showing different applications can be found in [47].

(30)

2.2 Permanent magnet synchronous machine

elec-tromagnetic model

Electrical machine voltage models can either be expressed using flux linkages Ψ or currents i as state variables. For synchronous machines it is convenient to express the model in the dq reference frame, synchronous with the rotor. The dynamic model of the PM synchronous machine in the dq reference frame, using flux linkages as state variables, can be written using the complex notation as [48]

v_dq= Rsidq+ dΨ_dq dt + jωΨdq (2.1) T = 3 2p Ψ_dq× idq (2.2) where × is the cross (or vector) product, Rs is the stator resistance, ω is the

electrical angular frequency and p is the pole-pair number. The model in (2.1) and (2.2) is non-linear. In fact, the flux linkages depend on the current components

Ψd= Ψd(id, iq) (2.3)

Ψq= Ψq(id, iq) (2.4)

It is often convenient to express the PM synchronous machine model using the currents as state variables. This because currents are often used in closed looped feedback control as these are much easier to measure with respect to flux linkages. The flux linkages can be expressed introducing the inductance matrix L and considering that by definition the PM rotor ΨPM flux is oriented on the d-axis

Ψ_dq= ΨPM(iq) + Lidq (2.5) L = " Ld(id, iq) Ldq(id, iq) Lqd(id, iq) Lq(id, iq) # . (2.6)

Note that the PM flux linkage in (2.5) is dependent on the q-axis current due to the iron saturation. The voltage equation (2.1) becomes

vdq= Rsidq+ dΨ_dq di_dq di_dq dt + jω ΨPM(iq) + Lidq . (2.7)

The flux linkage components partial derivatives are called incremental inductances [48] dΨ_dq didq = Linc = " ld(id, iq) ldq(id, iq) lqd(id, iq) lq(id, iq) # . (2.8)

(31)

2.3. COOLING OF TRACTION ELECTRIC MOTORS AND DRIVE-TRAINS

Unless the machine is linear Linc6= L. For control purpose, as well as for modelling

of the torque output, it is often an acceptable approximation to consider Linc = L

and simplify the inductance matrix to

L ≈ " Ld(id, iq) 0 0 Lq(id, iq) # . (2.9)

The two inductance components Ld(id, iq) and Lq(id, iq) can then be mapped

using numerical methods and used in the models. The torque equation can now be expressed as T = 3 2p h ΨPM(iq)iq+ (Ld(id, iq) − Lq(id, iq))idiq i (2.10) The first component of the torque is generated by the interaction of the PM flux and the stator current in the q-axis. The second component of the torque, also called reluctance torque, is generated by the saliency of the machine ξ = Lq/Ld. In

isotropic machines Ld= Lq and ξ = 1, while for IPM machines Ld< Lq meaning

that by utilizing a negative d-axis current both torque components are present and contributing to the output torque.

2.3 Cooling of traction electric motors and

drive-trains

Traction electric systems are designed for high power density and reliability. In order to achieve this, an effective cooling system for the electrical machine is needed, typically a closed loop forced liquid cooling. The main forced cooling solutions found in automotive and rail industry can be categorized as [19, 20]:

• Cooling jacket (oil or water) • Hollow shaft (oil or water) • Direct winding cooling (oil) • Fluid bath (oil)

• Fluid spray (oil)

Some of these are compatible and applied in the same machine.

Depending on the solution in use, there are different options when it comes to the coolant fluid, as shown in [20]. Most solutions use either oil or water (either pure water or with glycol). The main characteristics of these two coolants are

(32)

• Oil has a higher boiling point than water, so it can be used to cool the machine even at a temperature of 100◦C or higher. However, water-cooling may also exceed 100◦C if pressurized or wixed with a certain glycol percentage content.

• Oil is an electrical insulator, thus it can be in direct contact with the winding. Also, a leak from the cooling channel to the airgap would not cause any hazard. If coolant water should similarly leak, substantial machine damage might occur.

• Oil is already present as a lubricant in the transmission and naturally helps to prevent corrosion.

• Oil has a viscosity which is higher (and with a significant temperature dependence) compared to water, meaning that more pumping power is needed to circulate the fluid.

• The specific heat of water or water/glycol is about twice that of oil, so a given flow rate of water absorbs more heat per degree increase in temperature than the same flow rate of oil.

2.4 Thermal modelling of electrical machines

Depending on the phenomena to be observed and the level of accuracy required, the modelling of electrical machines can be done at different complexity levels. The main issue in modelling a fully coupled electromagnetic and thermal problem is the different time constants. The thermal time constant can be many orders of magnitude higher than the electrical one. A common approach to multiphysics modelling of electrical machines is to solve the electromagnetic model separately from the thermal model and eventually iterate.

Electrical machine thermal analysis can be divided into two basic methods: analytical lumped-parameter (LP) circuit and numerical methods. Although the LP approach has the speed advantage, the main effort is in the determination of the parameters of the circuit model [49, 50] and usually it is limited to well known geometries. With numerical analysis, computational fluid dynamics (CFD) and finite element analysis (FEA), any machine geometry can be modeled. How-ever, this can be demanding in terms of model setup and computational effort. Often numerical methods are used for the determination of the parameters in the analytical model [50, 51]. There are several possible approaches to thermal

(33)

2.4. THERMAL MODELLING OF ELECTRICAL MACHINES

modelling:

1. LP thermal model based on geometry, material data and empirical formulas for convection heat transfer coefficients

2. Thermal FEA with empirical formulas for convection heat transfer coefficients 3. LP with partial FEA and/or CFD, where some critical parts of the machines are modelled to increase accuracy. This is sometimes referred to as co-simulation

4. Computationally efficient model of liquid cooling coupled with FEA thermal modelling, presented in Paper V

5. Coupled 3D CFD and thermal FEA with iterations between the physics, presented in Paper IV

The list presented is created based on increasing computational effort and/or complexity of the machine/cooling geometry to be modelled. The phenomena of interest can differ among the models. With increasing complexity, more details can be observed. For example, the velocity distribution within the cooling channels or the distributed heat transfer coefficient on the walls of the cooling channels can only be determined by performing a CFD.

Furthermore, when implementing real-time observers or running extensive power-train optimization algorithms, the thermal model implemented needs to be reduced even to a lower level of computational effort. Two methods found in literature to do this are:

1. Eigenmodes simulation of the thermal dynamics 2. Reduced LP thermal model

The first method is based on the derivation of the eigenmodes from a full 3D thermal simulation. An example of this implementation for a traction electric machine has been presented recently in [52]. The second method is to reduce the LP network to the minimum number of nodes. An implementation of node reduction is presented in [53], showing a reduction from seven to three nodes without losing accuracy in the point of interest.

2.4.1 Lumped-parameter network - thermal parameters

In a lumped-parameter thermal model, the different parts of the electrical machine or inverter are treated as lumps assuming spatially uniform material properties

(34)

between nodes and a constant temperature in the part represented by each node [54]. The thermal capacitance of a node, modelling a region with n solids of different materials, assuming uniform heat capacity of the materials can be found as

Cth = n

X

i=1

Viρm,iCp,i (2.11)

where ρmand Cpare respectively the specific mass and specific heat of the material,

and V is the volume. The thermal resistance due to heat conduction along a line l for a region with varying area A perpendicular to the direction of the heat flow can be calculated as Rth = Z l 0 1 λA(x)dx (2.12)

where λ is the thermal conductivity. If the area is constant along the line this integral becomes

Rth=

l

λA. (2.13)

While for a segment of a hollow cylinder, the thermal resistance in the radial direction, (2.12) becomes Rth= ln rout rin λθlcyl (2.14) where routand rinare respectively the outer and inner radius of the hollow cylinder

segment, θ is the segment angle and lcylis the length of the cylinder. The thermal

resistance due to heat convection is calculated as Rth=

1

hA (2.15)

where h is the convection heat transfer coefficient. More details about heat transfer modelling can be found in [55].

2.5 The calorimetric measurement method

Calorimetry is a recognized means for the direct measurement of losses in liquid cooled power electronics and electrical machines. It can be used to overcome difficulties related to measuring a small relative difference between input and output power by a direct loss measurement. For a generic device under test (DUT) the losses can be directly estimated by measuring the mass flow rate and the temperature of the liquid at the inlet and outlet of the DUT using the simplified steady-flow thermal energy equation

(35)

2.5. THE CALORIMETRIC MEASUREMENT METHOD

where Cp is the heat capacity at constant pressure of the coolant, ˙V is the

volumetric flow rate, ρmis the specific mass of the fluid and ∆Θ is the temperature

difference between the inlet and the outlet. Note that Cpcan have a significant

temperature dependence for fluids such as oils.

It is of great importance to minimize the heat leakage of the DUT to the ambient to get an acceptable accuracy in the measurements. This is usually done for power electronic converters by using closed calorimetric boxes with internal controlled temperature and for the electrical machine by thermally isolating the machine frame and flange.

(36)

(37)

Design of the tooth coil

wind-ing traction PM machine

This chapter covers the sizing and design of the traction electrical machine for a passenger car. Starting from the vehicle description, the design specifications of the traction machine are defined and used for sizing. Then both electromagnetic and thermal design are analyzed, describing the main choices and trade-offs.

3.1 Vehicle description and design specifications

In this section, the design specifications for the traction machine are derived from the vehicle performance requirements. The reference vehicle is a 2-seater small city car with the specifications and performance requirements shown in Table 3.1. A fixed gear ratio between the axis and the electrical machine is assumed. The gearbox gear ratio is calculated as G = 8.4 assuming a base speed of the motor of 3600 rpm at 50 km/h vehicle speed and a maximum allowable motor speed ωmax of 11000 rpm at 150 km/h vehicle speed. The machine needs to be sized

to fulfill the three requirements listed in Table 3.1 in terms of acceleration and gradability considering an additional mass of 200 kg (two passengers plus luggage). The three requirements are represented in Fig. 3.1, in red, together with the machine outcome of the sizing process which is also converted into force at the wheels and achievable vehicle speed, in blue. This is showing how all the required operating conditions can be covered by the proposed machine size. The green curve presents a conservative way to calculate the 0-100km/h acceleration time with three different acceleration steps, each lasting 4 s.

(38)

Chapter 3 Design of the tooth coil winding traction PM machine

Table 3.1: Vehicle performance requirements

Quantity Symbol Value

Curb mass mc 800 kg

Front cross sectional area Avh 2 m2

Top speed vmax 150 km/h

Aerodynamic drag coeff. Cd 0.3

Rolling resistance coeff. Cr 0.009

Wheel radius rw 0.31 m

Starting gradability - 25 %

Hill climbing ability - 90 km/h at 6% and 1 m/s2

Acceleration 0-100 km/h - 12 s 0 20 40 60 80 100 120 140 160 Vehicle Speed [km/h] 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Force [N] no gradient, 1 m/s2 no gradient, 1.5 m/s2 no gradient, 2 m/s2 no gradient, 3 m/s2 no gradient, 3.5 m/s2 6% gradient 0-100 km/h 25% gradient 4s 4s 4s

Figure 3.1: Vehicle performance requirements in red and electrical machine sizing in blue. The green curve represents the 12 s acceleration to reach 100 km/h.

Standard laminated steel materials, such as M235-35, result in reasonable core losses with an excitation frequency of 0.1-1.0 kHz. Furthermore, to have a low current ripple, the switching frequency is assumed to be at least 20 times the fundamental. For IGBT converters, the switching losses of the converter should be

(39)

3.2. MACHINE DESIGN

limited by limiting the converter switching frequency to maximum 20 kHz. The switching frequency can be higher if MOSFETs are used but it is still important to keep the fundamental within a reasonable frequency to limit core losses. For the reasons above, the maximum fundamental frequency is set to ff,max= 1 kHz.

The maximum number of pole pairs can then be calculated as pmax=

2πff,max

ωmax

(3.1) resulting in 5.4, meaning that the maximum number of pole pairs is set to 5. The EM input design specifications are listed in Table 3.2. As a conservative approach, thanks to the intended high performance cooling, the machine is sized such that it can withstand the peak conditions for 30 seconds, assuming that its starting winding temperature is 100◦C. The value for maximum temperature of the coolant in Table 3.2 is assumed based on typical values found in industry. The maximum winding temperature is set assuming a class H insulation and the dc bus voltage of 600 V is based on typically used electric vehicles battery voltages between (250 V and 800 V).

Table 3.2: Electrical machine design specifications

Quantity Symbol Value Unit

Peak torque 30 s Tmax 140 Nm

Peak power 30 s Pmax 60 kW

Base speed ωb 3600 rpm

Max speed ωmax 11000 rpm

Coolant max Temperature Θmax,c 60 ◦C

Max winding Temperature Θmax 180 ◦C

DC bus voltage Vdc 600 V

As a reference, a similar small city car such as the new Smart Fourtwo, which has a curb mass of 880 kg and is available fully electric, mounts a 66 kW synchronous machine.

3.2 Machine design

The sizing procedure used for the main investigated motor is outlined in Paper III. The rotor has been chosen as an internal V-shaped PM rotor with air barriers similar to the Toyota Prius electric motor. The reasons for this choice are mainly

(40)

the saliency which improves the field weakening characteristic, the mechanical robustness of the solution and the lower PM losses, compared to a rotor surface magnet solution, thanks to the iron between the magnets and the airgap.

The coil disposition and geometries of the stator and rotor with details about disposition of conductors and cooling channels are shown in Figure 3.2. The 12 slot 10 pole machine has a key winding factor of 2 [25], meaning that each phase coil consists of two electrically series connected coils on adjacent teeth. Each coil has 28 turns, which allows for a 1.6 mm diameter enameled copper wire to be used. Having a bobbin which can be inserted, limiting the conductor diameter and avoiding parallel strands, enables the use of a linear winding machine, which could significantly reduce the manufacturing cost at high volume production. Each set of two series coils are then parallel connected to form a full phase winding.

Figure 3.2: Details of lamination geometry, cooling channels and conductor disposition. Left: stator and rotor laminations geometry, coil disposition and cooling channels. Right: the arrangement of copper conductors, potting material and oil cooling channel within one slot.

The manufacturing aspects and details of the machine prototype are pre-sented in Paper X together with some of the design improvements that could enable a mass manufactured product.

3.2.1 Electromagnetic Design

The electromagnetic design of the machine is covered in Paper IX and the choice of the pole slot combination is based on the study performed in Paper II. Additional

(41)

3.2. MACHINE DESIGN

aspects not found in Paper IX are discussed here, in particular, field weakening properties, stator winding AC losses and demagnetization. Efficiency and loss results presented in Paper IX are consequently adjusted, accounting for AC losses in conductors which are shown to play a substantial role at high speeds.

Yoke Cooling channel size and electromagnetic effects

This section is extracted from Paper VII and Paper IX. Cooling channels in the stator back yoke can be introduced by removing iron after the split ratio optimisation. However, if the slot and pole combination isn’t chosen properly, the yoke cooling channels might affect the electromagnetic performance negatively by increasing the reluctance path for the rotor PM flux and/or the linked flux between stator coils. The Q12p10 machine is a good choice since it features low mutual inductance between phases by linking the vast majority of the flux generated by phase windings in a loop contained within the two adjacent teeth belonging to the same phase group. This scenario is illustrated in the top part of Fig. 3.3, without the remanence PM flux and with 100 A in phase A. The change in self-linked flux due to magnetic saturation, caused by the cooling channels positioned between the phases, is negligible when channels are sized at hbar=2.0 mm. Fig. 3.3 shows

that the flux density in the yoke with barriers is still well below saturation values. To find out the performance implications of flux barrier between the phases, including the PM flux, a parametric sweep FEA has been performed. The cooling channels height hbar is swept from from 0.1 to 6.0 mm, corresponding to cooling

channels occupying in total 2-92% of the stator back yoke height. The final design is using a 2.0 mm barrier height, as shown in Fig. 3.3. Average torque, torque ripple and iron losses are evaluated at 5000 rpm, from zero up to rated peak current; 140 A RMS (35 A/mm2_{). The results for the highest current, which has}

the most dramatic impact, are shown in Table 3.3. The average torque output is monotonically decreasing with increasing barrier height. However, more than half of the yoke height (4.0 mm) can be cut out before 1% average torque loss is experienced. The relative torque ripple (pk2pk/avg) is kept robustly at 8% for all barrier sizes except the largest value. Moreover, the PM flux percentage reduction due to the introduction of barriers is negligible, also presented in Fig. 3.3. Regarding iron losses, when seen relative to the mechanical output power (Piron/Pmech), a small monotonically increasing trend can be seen of less than 0.1

unit of percent from the smallest barrier to the largest. Using dual 2 mm coolant barriers positioned between the phase groups is considered to have negligible impact on electromagnetic performance, and still offer enough cross-section area

(42)

Figure 3.3: FEA established magnetic flux generated by 100 A in phase A with (bottom) and without (top) remanence flux in the magnets. hbar=2 mm

for low-viscosity oil to flow without significant pressure drop. The shape of the channels and the choice of having four parallel channels instead of one is driven by flow split evaluation as presented in Paper VII. The resulting pressure drop is measured to 37.6 kPa at the maximum oil flow of 6.0 l/min at room temperature.

This type of utilization of part of the stator yoke to introduce cooling channels can be generalized for all TCWMs featuring a key winding factor equal to an even integer; which corresponds to an even number of adjacent tooth-coils belonging to the same phase winding as described in Paper II.

(43)

3.2. MACHINE DESIGN

Table 3.3: Results from FEA evaluation of cooling barrier size at 5000 rpm, 140 A RMS (35 A/mm2), 120◦current angle. hbar=0.1 mm is selected as reference

for relative changes. hbar=2.0 mm is chosen for the prototype machine.

Barrier height Average torque Torque ripple Iron losses PM Flux

hbar 2hbar/hyoke pk2pk pk2pk/Avg Piron/Pmech change

(mm) (%) (Nm) (± %) (Nm) (%) (W) (%) (± %) 0.1 1.5 145.66 0.00 12.34 8.47 980.2 1.29 0.00 1.0 15 145.62 -0.03 12.51 8.59 986.5 1.29 0.00 2.0 31 145.59 -0.05 12.24 8.41 999.0 1.31 0.00 3.0 46 145.35 -0.21 12.25 8.43 1009.5 1.33 -0.02 4.0 62 144.00 -1.14 12.03 8.35 1021.3 1.35 -0.17 5.0 77 140.04 -3.86 11.43 8.16 1006.8 1.37 -0.58 6.0 92 132.34 -9.14 14.24 10.76 958.9 1.38 -1.13 AC losses in conductors

There are several phenomena that cause additional losses in electrical machine windings [56]. These are:

• skin effect caused by current in the conductor itself • proximity effect

• skin effect caused by an external-field

• winding element circulating currents in multi-strand conductor bunches • phase winding circulating currents between parallel winding elements of the

same phase

For the machine analyzed in this work, skin effect due to the current in each conductor itself can be neglected because the skin dept at the maximum operating fundamental frequency, calculated (combining Ampere’s Law, Faraday’s Law and Ohm’s Law) as

δ =

r _ρ

πff,maxµrµ0

, (3.2)

where ρ and µr are respectively the resistivity and relative permeability of copper,

µ0 is the vacuum permeability and ff,maxis the maximum fundamental frequency.

The value obtained for the machine presented in this chapter is 2 mm, which is more than twice the radius of the conductor (0.8 mm). The conductors in the prototyped machine are not multi-stranded, so circulating currents among the strands do not occur. Phase winding circulating currents between parallel windings of the same phase could occur into the machine because of the parallel connection described in Paper IX. The cause of these circulating currents is usually some

(44)

magnetic unbalance in one of the parallel elements that induces an EMF different from the other branch. Causes of this magnetic unbalance could for example be: eccentricities, demagnetization, manufacturing differences of coils and magnets [56]. These sources of magnetic unbalance can not be predicted during the design phase so this effect is also neglected.

The two main sources of additional AC losses in the windings considered for the machine presented in this work are proximity effect and skin effect caused by an external-field. This external field in a PM machine is typically caused either by the leakage flux in the slot generated by the winding itself or by the rotor magnets through the slot opening. The latter can be even more relevant for a machine without tooth tips as the one designed in this work.

A 2D FEM model of the machine is made to evaluate the AC losses, modelling each conductor in one of the slots and accounting for eddy effects. The model is built with a dense enough mesh to capture the field change in different parts of each conductor. Fig. 3.4 shows the current density due to eddy currents distribution caused by different sources with and without stator current excitation. The no load losses generated in the windings are significant at high speeds.

Figure 3.4: Left: current density distribution of eddy currents caused by the rotor PMs flux through the slot opening at 5000 rpm. Right: current density

distribution of eddy currents caused by proximity effect and slot leakage flux at 5000 rpm and 105 A RMS.

In order to try and reduce the effect of the PM flux through the slot, the introduction of magnetic slot wedges with a relative permeability of 5 is evaluated

(45)

3.2. MACHINE DESIGN

with FEA. The results for no-load losses and AC to DC resistance ratio, at different currents, are presented in Fig. 3.5. The magnetic wedge plays a role in reducing the no-load losses but does not affect considerably the increase in AC resistance when current is flowing. Approximately a 30 percent reduction in no-load losses and a 3 to 4 percent reduction in losses during load is possible. This is due to that the main source of resistance increase comes from the proximity effect and not from the PM field through the slot opening. Note that no magnetic wedges have been installed on the prototype. The AC winding losses are found to be significant

0 2000 4000 6000 8000 10000 12000 speed [rpm] 0 100 200 300 400 500 600 losses [W] 0 A, no wedge 0 A, r=5 wedge 0 2000 4000 6000 8000 10000 12000 speed [rpm] 1 2 3 4 5 6 R AC / R DC 70 A, no wedge 70 A, r=5 wedge 140 A, no wedge 140 A, r=5 wedge

Figure 3.5: Left: AC losses in the winding at no load for the prototype design (no wedge) and with the addition of a magnetic wedge. Right: increase in resistance due to AC losses, mainly proximity effect.

and need to be modelled and included in the loss and efficiency maps. In order to do so these have been modelled analytically. The relation between AC resistance increase and speed has been found to be quadratic while inversely proportional with current. The interpolation with speed and current of the AC resistance is shown in Fig. 3.6. The fitted equation is

RAC RDC = 1 + kfit,I I0 I n2_rpm (3.3)

where the fitting coefficient for I0 = 70 A is kfit,70 = 3.8610−8 1/rpm2. This

fitting function has been utilized in the FEA loss map results to obtain updated values. Also the AC winding losses at no load have been fitted and adjusted to be included in the loss and efficiency maps. Note that in Paper IX, the no load losses have been measured, however the effect of AC winding losses has been neglected and these have been included in the iron losses. A scaling factor of 2 was applied to match the iron losses found in the FEA maps. When considering

(46)

the AC winding losses this scaling factor is reduced to 1.8 so also the iron losses are slightly re-scaled here. The updated loss maps and efficiency map including the effect of AC winding losses are presented in Fig. 3.7 and Fig. 3.8. Results show how copper losses are now speed dependent and the efficiency falls below 90% in the region above 9000 rpm.

0 2000 4000 6000 8000 10000 12000 speed [rpm] 0 2 4 6 8 10 12 R AC / R DC 70 A 140 A 70 A fitted 140 A fitted 35 A fitted

Figure 3.6: Fitting of AC resistance to speed and current.

Total losses [W] 250 500 500 750 750 1000 1000 1250 1250 1500 1500 1500 2000 2000 2000 2500 2500 2500 3000 3000 3000 4000 40005000 5000 0 2000 4000 6000 8000 10000 0 50 100 150 Torque[Nm] Copper losses [W] 250 250 500 500 750 750 750 1000 1000 1000 1250 1250 1250 1500 1500 1500 2000 2000 2000 2500 2500 2500 3000 3000 3000 4000 0 2000 4000 6000 8000 10000 0 50 100 150 Iron losses [W] 50 50 100 100 150 150 200 200 300 300 400 400 500 500 600 600 700 700 800 800 900 900 1000 ₁₂₅₀ 0 2000 4000 6000 8000 10000 speed[rpm] 0 50 100 150 Torque[Nm] PM losses [W] 20 20 20 40 40 40 60 60 60 80 80 80 100 100 150 150 200 200 250 250 300350 0 2000 4000 6000 8000 10000 speed[rpm] 0 50 100 150

Figure 3.7: Updated loss maps including AC winding losses (relative to Paper IX ).

(47)

3.2. MACHINE DESIGN 30 30 30 30 30 50 50 50 50 50 70 70 70 70 70 75 75 75 75 75 80 80 80 80 80 85 85 85 85 85 87 87 87 87 87 89 89 89 89 89 89 91 91 91 91 91 91 92 92 92 92 92 92 93 93 93 93 93 94 94 0 2000 4000 6000 8000 10000 speed[rpm] 0 20 40 60 80 100 120 140 Torque[Nm] A C E F D B

Figure 3.8: Updated efficiency maps including AC winding losses (relative to Paper IX ). The letters in red represent the six points where the efficiency has been validated, results are reported in Chapter 4.

Field weakening and characteristic current

One important aspect for traction machines is the field weakening characteristic, in particular the maximum output power-curve in the speed region above base speed. These aspects are directly related to the machine parameters. The relevant parameters are the d and q axis inductances, the PM flux linkage, the rated voltage and maximum current. The inductances and the PM flux linkage, seen in Fig. 3.9, are not constant with the currents, and this should be accounted for when analyzing the field weakening properties, as shown in Paper IX. The characteristic current of the machine [57], defined as the negative d-axis current needed to completely oppose the PM flux linkage, can be calculated as

Ich=

ΨPM

Ld

. (3.4)

Assuming that the machine is operating close to the maximum current of 140 A RMS, with the optimal maximum torque per ampere (MTPA) current angle, the param-eters of the machine can be fixed to constant values, Ld = 0.9 mH, Lq = 1.35 mH

and ΨPM= 0.095 Wb. The resulting characteristic current is Ich= 105 A, lower

than the maximum current, which, according to [57], means that the machine can be completely de-fluxed and in turn reach high speeds without exceeding the

(48)

Chapter 3 Design of the tooth coil winding traction PM machine 0 20 40 60 80 100 120 140 160 180 iq[A] 0.085 0.09 0.095 0.1 0.105 PM [Vs]

Figure 3.9: FEA results for PM flux linkage as a function of q-axis current.

rated voltage. These values can be used to analyze the machine inverter utiliza-tion on the IPM optimal design plane as a funcutiliza-tion of saliency and normalized PM flux and the maximum output power as a function of speed [57] shown in Fig. 3.10. The optimal machine design area, from a constant power speed curve perspective, lays in-between the 0.6 and 0.8 inverter utilization lines in the right plot of Fig. 3.10. The values of inverter utilization are closely related to the power factor. For low values of inverter utilization the power factor of the machine is poor, while for values close to 1 it is very hard to deflux the machine at high speed. The built machine lies close to the optimal machine design area and overall the machine constant power speed range fulfills the required specifications. Note that in Paper VII and Paper IX the machine peak power was reported to be 50 kW by considering the peak torque at base speed (3600 rpm), however this has been modified in this thesis to report the overall peak power of 60 kW which can be achieved at a speed which is slightly above 5000 rpm.

In Paper VI a machine with the same geometry as the prototype has been evaluated over the field weakening region with different control strategies. The two control strategies evaluated are minimum loss control (MLC) and MTPA. Note that the machine in Paper VI has a different voltage rating (400 V), which also meant a lower number of turns, magnets with a lower remanence flux and a higher current rating thanks to an increased diameter of each coil conductor (or several strands in parallel with a total equivalent conductor area which is increased compared to the built machine). Still most of the conclusions are applicable to the prototype built. The results presented in Paper VI do not account for AC losses in the windings which can reduce substantially the advantage of implementing the minimum loss control strategy. However, assuming that the AC losses are minimized during the design process by having multiple strands in parallel for each

(49)

3.2. MACHINE DESIGN

Figure 3.10: Left: maximum output power as a function of speed. Right: inverter utilization on the IPM optimal design plane as a function of saliency and normalized PM flux (black dot shows the machine in this work). [57].

conductor, the results, which show a relevant improvement in efficiency (above 10%) using MLC, are still valid.

Demagnetization

The machine should be able to withstand a converter fault without being damaged permanently. A severe condition in which the magnets could be permanently demagnetized is when the maximum current (140 A RMS) is set on the negative d-axis. In order to investigate, at the maximum current, whether the PMs are at risk of being demagnetized, a flux density plot set with the maximum amplitude to 0.55 T is shown in Fig. 3.11. The results indicate that the PM flux density does not fall below 0.55 T (magnets are of the magenta color in Fig. 3.11). According to the N48H Neodymium-Iron-Boron Magnets BH and JH curves from the manufacturer’s cut sheet [58], the magnets would not get demagnetized for this fault condition, up to a temperature of 120◦C, because the knee of the curve is at around 0.45 T.

The three-phase short-circuit fault at different operating points should also be investigated when studying the demagnetization. However, this is a transient condition and this verification is not performed at this stage.